* Drop Django 1.5 style {% url %} tag support
* Provide view name references for reverse etc
* Improve DjangoFQNamesProvider API to support "package/name only" check
It is not actually exposed anywhere in Python code style settings and
defaults to false thus affecting parenthesized multiline (chained)
qualified references and calls. Previously, oddly enough, this
behavior was controlled by ALIGN_COLLECTIONS_AND_COMPREHENSIONS flag.
It looks like this option is useless for these actions. It isn't available for Test run configurations - the most frequent use case for coverage. An output console isn't available for Concurrency Diagram, so users can't type any commands there. It doesn't work correctly with profiler's wrapping.
First, on matching type var with a union receiver type.
Though we're not able to match types like "Union[Type[A], Type[B]]" with
parametrized "Type[T]" to infer that T is Union[A, B] yet (should we?).
Second, on calling an annotated instance method on a class object
(when an argument for "self" is provided explicitly).
Its hard to check it with trial because for trial we use tempdir in tests and it fools rerun action to think testspect is set.
Changing this may be hard, it does not make any sense nor does
it affect production code. Leave it alone
- Don't clear the cache of packages regardless of whether
"pypi.python.org" is enabled or not (achieved along with fixes for
PY-24412)
- Properly hide PyPI packages in "Available Packages" if PyPI is
disabled
- Don't update PyPI cache on startup if it's disabled and don't
touch the respective timestamp
So as not to load the whole cache of available packages on EDT just
to get a timestamp for its invalidation in PyPackagesUpdater.
One simplification I made is that we no longer store the latest versions
of these packages in persisted settings. It would to much of a hassle
to keep it in sync then, and overall it's not that beneficial since
we're not able to get both package names and their versions in one
request due to the latest changes in PyPI API.
Namely, don't use Scope to find an annotation if AST access is not
allowed.
Added tests on handling of top-level variable annotations and class
attributes annotations in other files. Supposedly, we don't resolve
to unqualified target expressions in other scopes when unstubbing is
forbidden (ordinary local variables are not preserved in PSI stubs).
We search for them only in the same scope where the corresponding target
expression (of for loop, with statement or LHS of assignment with tuple
unpacking) is defined itself.
Another limitation is that this lookup is performed without proper
caching of this information on an affected target yet.
Additionally, I did the following:
* re-generated lists of supported/unsupported interpreter modules
* updated test data wherever Python versions appear in warnings
Still need to figure out, though, why _bz2/bz2 modules were not found
for some old versions of Python, hence information about their presense
should be updated by hand in versions.xml.
Ilya.Kazakevich
Mikhail Golubev
Semyon Proshev
Elizaveta Shashkova
Ekaterina Tuzova
Alexey Kudravtsev
Aleksey Rostovskiy